2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * The IP forwarding functionality.
11 * Many : Split from ip.c , see ip_input.c for
13 * Dave Gregorich : NULL ip_rt_put fix for multicast
15 * Jos Vos : Add call_out_firewall before sending,
16 * use output device for accounting.
17 * Jos Vos : Call forward firewall after routing
18 * (always use output device).
19 * Mike McLagan : Routing by source
22 #include <linux/types.h>
24 #include <linux/skbuff.h>
26 #include <linux/icmp.h>
27 #include <linux/netdevice.h>
28 #include <linux/slab.h>
34 #include <linux/tcp.h>
35 #include <linux/udp.h>
36 #include <linux/netfilter_ipv4.h>
37 #include <net/checksum.h>
38 #include <linux/route.h>
39 #include <net/route.h>
42 static bool ip_may_fragment(const struct sk_buff
*skb
)
44 return unlikely((ip_hdr(skb
)->frag_off
& htons(IP_DF
)) == 0) ||
48 static bool ip_exceeds_mtu(const struct sk_buff
*skb
, unsigned int mtu
)
50 if (skb
->len
<= mtu
|| skb
->local_df
)
53 if (skb_is_gso(skb
) && skb_gso_network_seglen(skb
) <= mtu
)
59 static bool ip_gso_exceeds_dst_mtu(const struct sk_buff
*skb
)
63 if (skb
->local_df
|| !skb_is_gso(skb
))
66 mtu
= ip_dst_mtu_maybe_forward(skb_dst(skb
), true);
68 /* if seglen > mtu, do software segmentation for IP fragmentation on
69 * output. DF bit cannot be set since ip_forward would have sent
72 return skb_gso_network_seglen(skb
) > mtu
;
75 /* called if GSO skb needs to be fragmented on forward */
76 static int ip_forward_finish_gso(struct sk_buff
*skb
)
78 struct dst_entry
*dst
= skb_dst(skb
);
79 netdev_features_t features
;
83 features
= netif_skb_dev_features(skb
, dst
->dev
);
84 segs
= skb_gso_segment(skb
, features
& ~NETIF_F_GSO_MASK
);
93 struct sk_buff
*nskb
= segs
->next
;
97 err
= dst_output(segs
);
107 static int ip_forward_finish(struct sk_buff
*skb
)
109 struct ip_options
*opt
= &(IPCB(skb
)->opt
);
111 IP_INC_STATS_BH(dev_net(skb_dst(skb
)->dev
), IPSTATS_MIB_OUTFORWDATAGRAMS
);
112 IP_ADD_STATS_BH(dev_net(skb_dst(skb
)->dev
), IPSTATS_MIB_OUTOCTETS
, skb
->len
);
114 if (unlikely(opt
->optlen
))
115 ip_forward_options(skb
);
117 if (ip_gso_exceeds_dst_mtu(skb
))
118 return ip_forward_finish_gso(skb
);
120 return dst_output(skb
);
123 int ip_forward(struct sk_buff
*skb
)
126 struct iphdr
*iph
; /* Our header */
127 struct rtable
*rt
; /* Route we use */
128 struct ip_options
*opt
= &(IPCB(skb
)->opt
);
130 /* that should never happen */
131 if (skb
->pkt_type
!= PACKET_HOST
)
134 if (skb_warn_if_lro(skb
))
137 if (!xfrm4_policy_check(NULL
, XFRM_POLICY_FWD
, skb
))
140 if (IPCB(skb
)->opt
.router_alert
&& ip_call_ra_chain(skb
))
141 return NET_RX_SUCCESS
;
143 skb_forward_csum(skb
);
146 * According to the RFC, we must first decrease the TTL field. If
147 * that reaches zero, we must reply an ICMP control message telling
148 * that the packet's lifetime expired.
150 if (ip_hdr(skb
)->ttl
<= 1)
153 if (!xfrm4_route_forward(skb
))
156 rt
= skb_rtable(skb
);
158 if (opt
->is_strictroute
&& rt
->rt_uses_gateway
)
161 IPCB(skb
)->flags
|= IPSKB_FORWARDED
;
162 mtu
= ip_dst_mtu_maybe_forward(&rt
->dst
, true);
163 if (!ip_may_fragment(skb
) && ip_exceeds_mtu(skb
, mtu
)) {
164 IP_INC_STATS(dev_net(rt
->dst
.dev
), IPSTATS_MIB_FRAGFAILS
);
165 icmp_send(skb
, ICMP_DEST_UNREACH
, ICMP_FRAG_NEEDED
,
170 /* We are about to mangle packet. Copy it! */
171 if (skb_cow(skb
, LL_RESERVED_SPACE(rt
->dst
.dev
)+rt
->dst
.header_len
))
175 /* Decrease ttl after skb cow done */
176 ip_decrease_ttl(iph
);
179 * We now generate an ICMP HOST REDIRECT giving the route
182 if (rt
->rt_flags
&RTCF_DOREDIRECT
&& !opt
->srr
&& !skb_sec_path(skb
))
183 ip_rt_send_redirect(skb
);
185 skb
->priority
= rt_tos2priority(iph
->tos
);
187 return NF_HOOK(NFPROTO_IPV4
, NF_INET_FORWARD
, skb
, skb
->dev
,
188 rt
->dst
.dev
, ip_forward_finish
);
192 * Strict routing permits no gatewaying
194 icmp_send(skb
, ICMP_DEST_UNREACH
, ICMP_SR_FAILED
, 0);
198 /* Tell the sender its packet died... */
199 IP_INC_STATS_BH(dev_net(skb_dst(skb
)->dev
), IPSTATS_MIB_INHDRERRORS
);
200 icmp_send(skb
, ICMP_TIME_EXCEEDED
, ICMP_EXC_TTL
, 0);